The processing pipeline is a set of instructions for the system how to process a certain video source. Usually you want to create one pipeline configuration per scene (location). If you have multiple files (sources) recorded from the same camera, you can process them all using this single pipeline.

List of pipelines

To add new pipeline click on Add new button.

Create pipeline config

First choose Task template from the pop-up menu. This template will automatically load process elements required for solving selected recognition task. Then specify the source that will be attached to this configuration and the name. If you leave the name without changes, the system will name it after the source’s name.

If you have a file with saved pipeline configuration, you can upload it back from here.

The pipeline page consists of following sections: Stack of video processing elements, Source with run/stop controls, Reference video frame, Reference meta-data frame and Advanced parameters.

Stack of processing elements

If you create your pipeline from the Task template, you’ll find a number of processing elements already added to your processing pipeline. Click RUN to see immediately how the system processes the video from your source.

The pipeline works as a stack of processing steps. Every frame from your source goes from top of the stack down to very bottom. Each of the processing elements in your stack adds/modifies information about the objects in the frame. If you click PAUSE – the system will show you a still frame with the recognized objects as well as the meta-data that it got from this frame.

Pipeline config

The processing pipeline is quite transparent: you can ‘inspect’ what one or another processing step does to your video frame or its meta-data by clicking on EYE on the header of each processing element. You can also use On/Off switcher to turn off the element temporary and see how this affected the recognition process.

The beginners usually just need to define recognition area and counting lines, whereas advanced users can create complex recognition scenarios by adding more processing elements to the pipeline.

Warning

If you want to add a Model or an Output, please ensure that you have 'initialized' it in the Models or Outputs section first.

Source

You RUN/STOP your recognition from this section. If the server is stopped, the run, pause, stop buttons are inactive. As soon as the server starts, these buttons become active.

  • start: The source reads the frames and send to the recognition pipeline.
  • pause/step forward: The source does not send any new frames to the recognition pipeline. Each click will feed a single frame to the pipeline and update reference meta-data section.
  • stop: The server stops processing this source.
  • Run at max-speed from Action drop-down menu: If you have 'Files' processing server (not 'live' streams) you can speed up the processing. (by default it runs at 25fps)
  • goto start: the system will start processing your file or batch of files from very beginning.
  • change source: You can attach another source of video to the pipeline.

process controls

Reference Video/Meta-data

If the recognition process is running, you'll be receiving life video. You can adjust Pipeline video resolution to reduce the load on the network or on the server that needs to perform additional video-compression before sending to you. This doesn't affect the resolution of the video stream that goes to the processing pipeline.

If the process is paused, you will get single image of original resolution together with corresponding meta-data.

Meta data JSON

Advanced parameters

Use advanced parameters section to adjust the processing behaviour:

  • Processing frequency: How many frames per second should the source generate for the processing pipeline.
  • OSD options: Use checkboxes in the OSD parameters group to switch on/off drawing of bounding boxes, tracks, object names, events etc.
  • Track smoothing: Increase value to see smooth tracking lines.
  • OSD lines thickness: How thick should be the bounding boxes and track lines.
  • Text size: You can increase/decrease OSD text size.
  • Performance logging: You can log how much time each processing step takes. This parameter allows you setup log frequency (in seconds).
  • Schedule: You can run this processing pipeline only within during time intervals.
  • Description: This is a user plain text memo.

Advanced recognition parameters

Important

You can configure and run as many pipelines as you need. Start/stop one or several pipeline and monitor their status from the list of pipeline configurations.